Analytical Modeling of Multipass Welding Process with …

Analytical Modeling of Multipass

R. N. S. Fassani
Universidade Estadual de Campinas
Departamento de Engenharia de Fabricação
Cx. Postal 6122
13083-970 -Campinas, SP. Brazil
sfroseli@uol.com.br
O. V. Trevisan
Universidade Estadual de Campinas
Departamento de Engenharia de Petróleo
Cx. Postal 6122
13083-970 Campinas, SP. Brazil
trevisan@dep.fem.unicamp.br
Welding Process with Distributed
Heat Source
In the welding process, the most interesting regions for heat transfer analysis are the
fusion zone (FZ) and the heat affected zone (HAZ), where high temperatures are reached.
These high temperature levels cause phase transformations and alterations in the
mechanical properties of the welded metal. The calculations to estimate the temperature
distribution in multiple pass welding is more complex than in the single pass processes,
due to superimposed thermal effects of one pass over the previous passes. In the present
work, a comparison is made between thermal cycles obtained from analytical models
regarding point (concentrated) and Gaussian (distributed) heat sources. The use of
distributed heat source prevents infinite temperatures values near the fusion zone. The
comparison shows that the thermal cycles obtained from the distributed heat source model
are more reliable than those obtained from the concentrated heat source model.
Keywords: Distributed source, analytical modeling, multipass welding

and each weld pass fulfills the whole etched groove (no heat flux in
Introduction the z direction).

Most of the published work on heat transfer during welding

processes considers that the heat source is concentrated in a very
small volume of the material. After such consideration, analytical
solutions are obtained assuming a point, a line or a plane heat z welding
source, as those proposed by Rosenthal (1941). However, direction
measurements of temperatures in the fusion and heat affected zones
differ significantly from the values provided by those solutions,
since the singularity located at the source origin results in infinite
temperature levels. These concentrated source models present higher O y
accuracy in regions where the temperature does not exceed twenty
percent of the material melting point (Goldak, Bibby and x
Chakravarti, 1984).1 Figure 1. Coordinate system used in the model.
In order to avoid the occurrence of unrealistic values at the
center and in the vicinity of the fusion zone (FZ), it is more
adequate to consider a distributed heat source in the model The formulation of the problem to the first weld pass is made up
development. In reality, the heat source is distributed in a finite by the one-dimensional transient heat conduction equation, and its
region of the material, a fact most relevant to the assessment of boundary and initial conditions. It is similar to the formulation of
temperatures near the FZ. There are several models for heat source the point heat source problem. In terms of θ (θ = T-To), it is:
distribution. The Gaussian distribution, firstly suggested by Pavelic
et al. (1969), is the most used. Although solutions considering ∂θ ∂ 2θ
distributed heat sources can be reached both analytically and =α 2 (1)
∂t ∂y
numerically, there is an increasing tendency to the use of numerical
methods. This work presents a new analytical solution to estimate
temperature fields in multipass welding, as generated by Gaussian θ ( t = 0) = 0 (2)
heat sources. The solutions were obtained from the known forms for
the multipass welding, for point heat sources. A case study, using θ( y → ∞ ) = 0 (3)
practical material and parameters, is also simulated to show the
main characteristics of the thermal cycles furnished by the ∞ Q
∫ θdy = ρc
1
developed model. A comparison with the results provided by the (4)
concentrated source corresponding solution is carried out. −∞

where:
Analytical Development T = temperature (oC)
In the one-dimensional model, the heat flux is considered to To = ambient temperature (oC)
occur only in the y direction, as shown in the coordinate system of θ = temperature difference (oC)
Fig. 1. The following assumptions are made: the heat source moves t = time (s)
at a sufficiently high speed (to neglect heat flux in the x direction), y = coordinate (m)
Q1 = thermal energy per unit area (J/m2)
α = thermal diffusivity of the material (m2/s)
ρ = density of the material(kg/m3)
Presented at COBEM 99 – 15th Brazilian Congress of Mechanical Engineering.
22-26 November 1999, São Paulo. SP. Brazil.
c = specific heat of the material (J/kgoC)
Paper accepted: July, 2002. Technical Editor: José Roberto de França Arruda.

302 / Vol. XXV, No. 3, July-September 2003 ABCM

 R. N. S. Fassani and O. V. Trevisan

The solution to this problem is known (Rosenthal, 1941), and it

3Q  3y 2 
is expressed by: q s ( y) = exp −  (10)
π y b Erf ( 3 )  y2 
Q1  y2   b 
θ( y, t ) = exp −  (5)
ρc 4παt  4αt 
  The diffusion process of an instantaneous Gaussian heat source
applied to the surface of the material may be obtained by the source
To take into account the distribution of the heat source, please method. Let the y coordinate, along which the heat source varies, be
refer to Fig. 2, where a source with normal or Gaussian distribution divided in small elements dy’. The heat dQ = qs(y’)dy’ is supplied to
is instantaneously applied at t = 0 to the surface of a plate. The the element dy’ at t = 0, and may be regarded as an instantaneous
center C of the source coincides with origin O of the coordinate point heat source. According to Eq. (5), the diffusion process to an
system xyz. The total power of the source is given by: instantaneous heat source is:

∞
dQ  y2 
Q= ∫ q s ( y)dy (6) dθ( y, t ) = exp − 
−∞ ρc 4παt  4αt 
 
where:
Q = total power of the source (W) or
qs = power of the source per unit length (W/m)
 d2 
q s ( y′)dy′
dθ( y, t ) = exp −  (11)
ρc 4παt  4αt 
y  
qmax
where d is the distance between the instantaneous source and a point
located on the y axis, that is,

d2 = (y-y’)2 (12)
C
Substituting Eqs. (10) and (12) in Eq. (11):
yb
 3y 2   2
dθ( y, t ) =
3Q
exp −  exp− ( y − y′)  (13)
πy b ρc 4αt Erf ( 3 )  y2  4αt 
 b  

By the superposition principle, the temperature change in the y

x point may be obtained by summing the contributions of all
Figure 2. Gaussian heat source. instantaneous concentrated sources dQ, acting along the y
coordinate of the material, between -yb and yb points:
In the one-dimensional case, the Gaussian distribution of the
 3 y ′2   ( y − y ′) 2 
b y
3Q
heat source along the y direction occurs simultaneously at all points
of the x direction of welding. The power qs (y) may be expressed by:
θ ( y, t ) = ∫  − yb2
πyb ρc 4α t Erf ( 3 ) − y b
exp  exp −
  4αt 
 dy ′

(14)
q s ( y) = q max exp(−Ay 2 ) (7)
Solving the integral and rearranging the solution, one obtains:
where:
qmax = qs maximum value (W/m) 3Q
θ ( y, t ) =
A = coefficient of arc concentration (1/m) 2ρc π(12 αt + y 2b ) Erf ( 3 )
Coefficient A is determined considering a distance yb in Eq. (7),    
exp  − y2
+
y 2 y 2b   12 αt − yy b + y 2b  
which corresponds to the distance from the origin to the location    Erf   +  (15)
 4αt 4α t (12 αt + y b ) 
2
 2 α t (12 αt + y 2 )  
where the power is reduced to five percent of its maximum value   b  
(Fig. 2). Thus,  
  

exp  − y 2 2 2
y yb   12 α t + yy b + y b  
2
3 +  Erf  
A= (8)   4αt 4α t (12 αt + y 2b )   2 α t (12 αt + y 2b )  
   
y 2b
Equation (15) is the solution to the first weld pass, regarding the
When yb is large, qs(y) decreases slowly with y. Substituting Eq.
input of a heat source with Gaussian distribution. The solution to the
(8) in Eq. (7) and then in Eq. (6), and integrating this equation
second pass is obtained from the point heat source solution
between -yb and yb limits, one obtains:
(Suzuki,1996):

3Q  d2   
q max = (9) qs1 (y′)dy′  + qs2 (y′)dy′ exp− d2
π y b Erf ( 3 ) dθ(y, t) = exp − 
 
ρc 4παt  4αt  ρc 4πα(t − t p )  4α(t − t p ) 
Equation (7) may then be written as: (16)

J. of the Braz. Soc. of Mech. Sci. & Eng. Copyright  2003 by ABCM July-September 2003, Vol. XXV, No. 3 / 303
Analytical Modeling of Multipass Welding Process with …

In Eq. (16), it may be observed that the t variable was displaced bead x width). Table 1 shows the physical properties used in the
by a value tp, which corresponds to the sum of the welding and simulation. It is known that the physical properties of the metal
waiting times to the beginning of the second pass. The use of indices change with temperature. However, this variation in the analytical
1 and 2 in the qs variable is for possible and sought variations of the models results in a non-linear equation, and it is not possible to
heat input between passes. The same steps applied to obtain Eq. (15) obtain the solution in closed form. Then, the physical properties are
are used to reach the solution for the second pass, and so on. usually taken at a specific temperature, for example, at half the
Analogously, the general solution to n passes, in terms of T, is given melting point of the material. In this work, they were calculated at
by: 800oC. The values refer to low carbon steels, but they can be used
for HSLA steel, as suggested by Hanz et al. (1989).
3 n Qi
T(y, t) = To + ∑ *
2ρc π i =1 12α[t − (i − 1)t ] + y2 Table 1. Physical properties of low carbon steels (Hanz et al,1989).
p b
k (J/msoC) ρc (J/m3oC) α (m2/s)
 y2 y2 y2b 
{exp− + 
31.67 7.14x106 4.44x10-6
 4α[t − (i − 1)t p 4α[t − (i − 1)t p{12α[t − (i − 1)t p ]}
Welding parameters. In order to fulfill the groove, in butt
 
 12α[t − (i − 1)t p ] − yyb + y2b  (17) welding, the usual practice is to increase the heat input from one
Erf + pass to the next. In the present simulation of a real case, the increase
 2 α[t − (i − 1)t p ]{12α[t − (i − 1)t p + yb} 
2
of heat input is obtained by increasing the welding current, the other
 
parameters in Eq. (18) remaining unaltered. However, the current
 y 2 2 2
y yb 
exp− + increase causes efficiency to decrease. Then, a different value of

 4α[t − (i − 1)t p 4α[t − (i − 1)t p{12α[t − (i − 1)t p ]} efficiency must be used in each pass. The choices of these values
were based on the efficiency range for the Gas Metal Arc Welding
 
 12α[t − (i − 1)t p ] + yyb + y2b  (GMAW) process, which ranges from 66 to 85% (Svensson, 1994).
Erf } The welding parameters used in the simulation are in Table 2. The
 2 α[t − (i − 1)t p ]{12α[t − (i − 1)t p + yb } 
2
heat input (HI) values were determined by Eq. (18), multiplied by
 
the material thickness. The same HI values were used in the point
Equation (17) is the solution to the temperature distribution in and Gaussian heat source models. However, in the Gaussian heat
one-dimensional multipass welding processes, supplied by Gaussian source model, only ninety-five percent of HI was applied to the
heat sources. Far from the heat source, i.e., for distances where y is weld. In order to adjust such difference, the HI values in Table 2
of the same magnitude as yb, Eq. (16) is similar to the solution were multiplied by 1.05 for the Gaussian model.
obtained for the point heat source. However, near the FZ and HAZ
(y<<yb), the correction introduced by the distributed heat source Table 2. Welding parameters used three weld passes.
approach in Eq. (17) allows to better predicting the temperatures in
Pass I (A) V (V) v (m/s) η (%) HI (x106
these regions.
J/m)
1 186 26.4 80 0.79
Model Evaluation 2 235 26.2 0.005 75 0.92
3 301 25.4 70 1.07
A comparison between concentrated and distributed heat source
models was made through simulation of thermal cycles for three Parameter yb. In order to verify the capability of the proposed
weld passes. Equation (17) was used to calculate the temperatures model to reproduce the thermal cycles, some values were chosen for
near the fusion zone, when a Gaussian distributed heat sources is the y and yb variables. These choices took into account the heat
applied. The multipass model with concentrated heat source is given input used in the simulation, and also known values of yb from the
by Eq. (16). In this case, the variable qs(y) does not have a
literature, obtained via experimental determination. Table 3 shows
distribution, and it is calculated by:
the yb values obtained by Kou and Wang (1986), Zacharia et al.
(1989), and Wu (1992), as well as the heat input (HI) used in their
ηVI
qs = Q = (18) analyses. In the present work, the yb parameter was estimated based
vδ on the heat input values showed in Table 3. In Eq. (8), the A
coefficient was determined for the yb distance, where the power is
where: reduced to five percent of its maximum value. According to this
η = arc efficiency (%) equation, by increasing the heat input parameter yb also increases,
V = welding voltage (V) such that the area under the curve of Fig. 2 remain equal to ninety-
I = welding current (A) nine percent of its qmax value. Therefore, different values of yb for
v = welding speed (m/s) each pass were considered, since the heat input increased in the
δ = plate thickness (m) second and third passes. The yb values used in the simulation were
0.004; 0.0047 and 0.0054 m for the first, second and third pass,
The waiting time (time between passes) used corresponds to 60 respectively.
seconds, and the welding process was simulated during a total time
of 300 seconds. The ambient temperature is 25oC. The error Table 3. Heat input and yb values in the literature.
function in Eq. (17) was evaluated using a polynomial
approximation. The properties and parameters used in the simulation Author I V (V) V HI yb (m)
are described below. (A) (m/s) (x106J/m)
Material. The evaluation of the proposed model was made Kou and Wang (1986) 100 11 0.0055 0.20 0.003
considering butt welding of high strength low alloy steel (HSLA) Zacharia et al. (1989) 175 14 0.0034 0.72 0.003
plates, with dimensions 0.13 x 0.10 x 0.25 m (thickness x length Wu (1992) 200 20 0.010 0.40 0.002

304 / Vol. XXV, No. 3, July-September 2003 ABCM


Results and Discussion
To compare the point and Gaussian heat source models, thermal
cycles were simulated at two different y locations, namely: y equal
to 0.001 m, appropriate for a position near the fusion zone (y < yb),
and y equal to 0.003 m, a location at a distance of the same
magnitude as the yb parameter.
Figure 3 shows the thermal cycles obtained after the point and
Gaussian heat source models, at y = 0.001 m. It can be observed that
the peak temperatures in each pass are higher for the point heat
source model than for the Gaussian heat source model. This occurs
due to the assumption that the heat input is instantaneously applied
over an infinitesimal volume cross-sectioned by the thickness-width
plane at the center of the workpiece. In the Gaussian heat source
model, it is assumed that the heat input is applied over the finite
volume, including the y coordinate at the extent of the yb parameter.
Therefore, the peak temperatures produced by the latter model is
expectedly more realistic. The previous determination of the peak
temperature to be reached at a specific location is interesting, since
it indicates fortuitous phase changes. The differences between the
temperatures simulated by the two models are described in the Table
4. T1 refers to maximum temperature reached in the first pass; T2 in
the second pass, and so on. It is worth noticing that the correction in
the proposed model affect only the peak temperature determination,
and no difference is seen in the cooling rate.
Table 4. Peak temperatures reached in each weld pass.
Tpeak Point heat source Gaussian heat source
T1 2057.52 1632.28
T2 2535.16 1881.01
T3 3044.18 2115.27
3500
3000
2500
2000 point
T (oC)
Gaussian
1500
1000
500
0
0 100 200
t (s)
Figure 3. Thermal cycles for the point and Gaussian heat source models,
at y = 0.001 m.
In Figure 4, the thermal cycles were simulated using y = 0.003
m. It is instrumental to show that, if y is close to yb, the Gaussian
heat source model is equivalent to the point heat source solution.
The two models provide results that are practically the same. This
means that the point source model at distances far from the fusion
zone can correctly predict the temperature fields.
J. of the Braz. Soc. of Mech. Sci. & Eng. Copyright  2003 by ABCM
R. N. S. Fassani and O. V. Trevisan
1500
1000
point
Gaussian
T (oC)
500
0
0 100 200 300
t (s)
Figure 4.- Thermal cycles for the point and Gaussian heat source models,
at y = 0.003 m.
Conclusions
The main conclusions of this work are:
- the closed form solution obtained allows to estimate the thermal
cycles produced by multipass welding process, near the fusion and
heat affected zones;
- the distributed heat source in the proposed solution is an important
correction for the known model with point source, since this factor
allows to obtain temperature values more realistic near to the fusion
zone;
- the analytical solution derived from the point source model can be
safely used to predict temperature fields away from the fusion zone
(FZ) and the heat affected zone(HAZ).
Acknowledgement
The authors want to acknowledge the financial support of
FAPESP - Fundação de Amparo à Pesquisa do Estado de São Paulo
- to the research work reported in the present paper.
References
Goldak, J., Bibby, M., Chakravarti, A., 1984, A new finite element
model for welding heat sources, Metallurgical Transactions B, vol. 15B, pp.
299 - 305.
Hanz, Z., Orozco, J., Indacochea, J. E., Chen, C. H., 1989, Resistance
spot welding : a heat transfer study, Welding Journal, vol. 68, n. 9, pp. 363s -
371s.
Kou, S., Wang, Y. H., 1986, Weld pool convection and its effect,
Welding Journal, vol. 65, n. 3, pp. 63s - 70s.
Pavelic, V., Tanbakuchi, R., Uyehara, O. A., Myers, P. S., 1969,
Experimental and computed temperature histories in Gas Tungsten Arc
300
Welding of thin plates, Welding Journal, vol. 48, n. 7, pp. 295s - 305s.
Rosenthal, D., 1941, Mathematical theory of heat distribution during
welding and cutting, Welding Journal, vol. 20, n. 5, pp. 220s - 234s.
Suzuki, R. N., 1996, Solução Analítica para Distribuição de
Temperatura no Processo de Soldagem MIG com Múltiplos Passes, Tese
Mestrado, Universidade Estadual de Campinas, Campinas, São Paulo, Brasil.
Svensson, L. E., 1994, Arc welding process, in Control of
Microstructures and Properties in Steel Arc Welds, CRC Press.
Wu, C. S., 1992, Computer simulation of three-dimensional convection
in traveling MIG weld pools, Engineering Computations, vol.9, pp. 529 -
537.
Zacharia, T., Eraslan, A. H., Aidun, D. K., David, S. A., 1989, Three-
dimensional transient model for arc welding process, Metallurgical
Transactions B, vol. 20B, pp. 645 - 659.
July-September 2003, Vol. XXV, No. 3 / 305